[Bug ld/31503] [hppa] Unsupported 14-bit PA 2.0 relocations for 32-bit (narrow) mode (elf32-hppa.c)

[dave.anglin at bell dot net]

https://sourceware.org/bugzilla/show_bug.cgi?id=31503

--- Comment #11 from dave.anglin at bell dot net ---
On 2024-03-28 7:34 p.m., dave.anglin at bell dot net wrote:
> https://sourceware.org/bugzilla/show_bug.cgi?id=31503
>
> --- Comment #10 from dave.anglin at bell dot net ---
> On 2024-03-28 7:20 p.m., amodra at gmail dot com wrote:
>> https://sourceware.org/bugzilla/show_bug.cgi?id=31503
>>
>> --- Comment #9 from Alan Modra <amodra at gmail dot com> ---
>> (In reply to Andreas Schwab from comment #7)
>>> That should use ALIGN(8).
>> ALIGN inside an output section statement doesn't align the section, just the
>> current *relative* value of dot.  So the effect is exacly the same as John's
>> expression.  They both align zero to a multiple of eight and thus do nothing.
> Okay.  I was hoping that the expression would align the value assigned to
> $global$.
>
> This change appears to successfully align .data:
>
> dave@mx3210:~/gnu/binutils/src/ld/scripttempl$ diff -u elf.sc elf32hppa.sc
> --- elf.sc      2024-03-28 21:45:32.560456976 +0000
> +++ elf32hppa.sc        2024-03-28 22:26:32.625611275 +0000
> @@ -669,7 +669,7 @@
>
>      ${DATA_PLT+${PLT_BEFORE_GOT-${PLT}}}
>
> -  .data         ${RELOCATING-0} :
> +  .data ${RELOCATING-0} ${CREATE_SHLIB-${CREATE_PIE-${RELOCATING+ALIGN(8)}}} 
> :
>      {
>        ${RELOCATING+${DATA_START_SYMBOLS}}
>        *(.data${RELOCATING+ .data.* .gnu.linkonce.d.*})
>
> But it causes two testsuite fails:
>
> PASS: Build pr23162b
> FAIL: Build libpr23161a.so
> PASS: Build pr23161a
> FAIL: Build libpr23161b.so
>
> regexp_diff match failure
> regexp "^Relocation section '\.rel(a|)\.dyn' at offset 0x[0-9a-f]+ contains
> [0-9
> ]+ entries:$"
> line   "Relocation section '.rela.got' at offset 0x19c contains 3 entries:"
> output is
> Relocation section '.rela.got' at offset 0x19c contains 3 entries:
>    Offset     Info    Type                Sym. Value  Symbol's Name + Addend
> 00001088  00000501 R_PARISC_DIR32         00001094   __bss_start + 0
> 0000108c  00000301 R_PARISC_DIR32         00001094   _edata + 0
> 00001090  00000401 R_PARISC_DIR32         00001094   _end + 0
>
> If this isn't serious, I can just modify the regexp.   Otherwise, do you have
> any thoughts
> on how to work around this issue.
Found it.  Need to define "GENERATE_COMBRELOC_SCRIPT=yes".

-- 
You are receiving this mail because:
You are on the CC list for the bug.